R&D Team Manager. Fujitsu Research India Pvt. Ltd. (APR-2022 to Now).
Role. Spearhead strategy, mentorship, and delivery for Generative AI, Agentic LLMs, Multi-Modal LLMs, Reinforcement Learning, and AI/ML R&D. Build and lead high-caliber teams; drive cross-domain innovation and business value. Drove innovations from patent to deployment, directly contributing to business strategy and new product lines.
Key Achievements (Invented & Filed 9 Patents + 3 Papers + enterprise solutions):
AI-Driven RCA (Root Cause Analysis) Pipeline for Log/Text Data. (Paper Accepted at ACL-2025 Findings, Titled: HG-InsightLog: Context Prioritization and Reduction for Question Answering with Non-Natural Language Construct Log Data).
Proactive RCA from Log/Text Data with Mamba and LLM.
Time series Forecasting enterprise solutions:
Explainable AI Solution: Explainable Timeseries forecasting and Anomalies Detection with LLMs & DL.
Forecasting Accuracy: Enhancing the training strategies for Multivariate Time Series Forecasting.
Spatiotemporal Multi-Step and Multi-Variate time series forecasting for 5G Network usage. (Paper published @ IEEE-Globecom-2023, ‘IEEE Flagship Conference).
Cold-Start Multivariate Time Series Forecasting.
Security based Solutions: Anomalies Detection in Timeseries Data (Paper published @ ACM Wisec-2024).
AI for Smart Infrastructure: Automated Real-Time Storm Chaser System.
The Causal Graph Linked Prediction for a large Gene Expression System.
Senior Research Staff and Manager- Samsung Research Institute, Bangalore (India), (May-2020 – to APR-2022).
Role: Managing, Mentoring, Consulting, and Conducting Research work in Conversational AI, NLP, Deep Learning, and Machine Learning.
Key Achievements:
Conversational AI based novel solutions:
The Scalable Multi-Intent System for Company’s conversational AI platform. (Paper published in IEEE Access).
Effective Solution for Out-of-domain Intent detection. (Paper published in IEEE Access).
Effective Solution for Missing Slot Detection Problem. ( US patent - (US 20230077874 A1).
Worked on an Emotional Intelligence supported Conversational AI system.
Age and Gender Prediction using Speech.
IoT accident prediction Solution: AI based Accident Prediction and End-User Support in Smart-Home environment. (US Patent - US20230402187A1 ) .
Scientist III (Senior Scientist), Computer Science & Eng. at Conduent Labs (Bangalore, India), (Apr-2018 to May-2020)
Role: (a) Leadership role, (b) bringing new research projects with strong business alignment, (c) mentoring team members, and (d) conducting high-level R&D works. Achievement (2019-2020)- Got Honor Roll Award from Conduent – For exceeding expectations by demonstrating outstanding attitude and excellent performance.
(Research Projects Details):
Automatic Crime Volume Prediction System. - (US Patent-US20200410321A1).
An Automatic Hate Target Identification System. (US Patent- US20210240938A1)
Automated Cyber Hate Profiling System. Achievements - (“Awarded by the Company”).
A Code-Mixed Language Model for Aggression Detection (published in COMAD-2020)
The adverse drug event detection system (an R&D project). Achievements - Won 1st prize in the Conduent’s 3i HUB (Innovate, Implement and Impact), Initiative.
Quantification of emotion and sentiment. (US Patent US20230325604A1).
Automated BOT-Humanization System for Conduent BOT-DARA. Achievements – A Conversational AI Research Project (Patent document Submitted).
Senior Machine Learning Scientist, Phenom People (Dec 2016 to Feb-2018)
Role: Led initiatives in social graph analytics for workforce planning, increasing business process efficiency and product adoption.
Key Projects:
Developed a novel weighted knowledge graph for (a) Smart B-2-B-2-C hiring, (b) Contextual and (c) Personalized Search.
Candidate Social Graph for automatic resource arrangements and supporting business planners and Automatic Job-Highlight System.
Post-Doctoral Researcher at University of California (Davis) (Sep-2015 to Sep-2016);
Worked on research project on automated AI-driven taxonomy extraction from software engineering documentation. Produced actionable frameworks for knowledge mining in engineering repositories.
Research Professional, TCS Innovation Lab, New Delhi (May 2013 to Aug-2015)
Research Project Details:
Developed and published award-winning systems for automatic plagiarism detection. (winner of the best paper award 1st place @ CICLing-2014)
Automatic text-quality grading system (Effectively grades the quality of E-mails, without relevant background model). (Paper published)
Automatic event detection, alignment and prediction system (related to economic events).
Research Intern; IBM IRL; Bangalore, India (May 2012 – July 2012) (Research Project details):
Research Project details: Developed advanced “Why-based” question answering using Wikipedia, improving natural language Q&A capabilities for enterprise search.
BEST PAPER AWARD (1st place) @ CICLING-2014 – Details: “Niraj Kumar; “A Graph-Based Automatic Plagiarism Detection Technique to Handle The Artificial Word Reordering and Paraphrasing”, A. Gelbukh (Ed.): CICLing 2014, LNCS 8404, pp. 481–494, 2014. (LINK)
IN TOP SYSTEM @ TAC-2011: My system was in the top system for “Automatic Summarization Evaluation Task” at Text Analysis Conference (TAC 2011), organized by National Institute of Standards and Technology (NIST), Gaithersburg, Maryland, USA TAC-2011. (For details, see My Publications)
IN TOP SYSTEM @ TAC-2010: My system was in the top system for “Automatic Summarization Evaluation Task” at Text Analysis Conference (TAC 2010), organized by National Institute of Standards and Technology (NIST), Gaithersburg, Maryland, USA TAC-2010. (For details, see My Publications).
Other Recognition: Our Unsupervised Phrase Identification technique for Keyphrase Extraction, has been appreciated by the survey paper published in COLING-2010, Titled: “Conundrums in Unsupervised Keyphrase Extraction: Making Sense of the State-of-the-Art”
Program Committee Member: (1) CODS-COMAD – 2018, 2019, 2020, 2021, 2022 (2) ICON 2013, 2016, 2020, 2021, (3) AI-ML Systems - 2022
Reviewed Journal paper: Oxford Journals -> Science & Mathematics -> “Computer Journal”, IEEE ACCESS.
I obtained my PhD, (CSE) from IIIT-Hyderabad (June-2015), under the guidance of Dr. Kannan Srinathan and Dr. Vasudeva Varma.
PhD Thesis Title: Towards Intelligent Text Mining: Under Limited Linguistic Resources.
Academic Qualification Summary:
1. PhD (CSE) IIIT-Hyderabad – July-2015
2. MCA – IGNOU (Delhi) – 2002 Jan - June-2004
3. BCA – IGNOU (Delhi) –1999-Jan – 2002-June
US20230402187A1: Method and system for mitigating physical risks in an IoT environment (Samsung, 2023)
US20230077874A1: Methods and systems for determining missing slots in voice interaction (Samsung, 2023)
US20200410321A1: Neural network systems and methods for event parameter determination (Conduent, 2020)
US20230325604A1: Method and system for automated sentiment classification (Conduent, 2023)
US20210240938A1: Neural network systems and methods for target identification from text (Conduent, 2021)
US20250173564A1: Spatiotemporal Multi-Step and Multi-Variate time series forecasting for 5G Network usage (Fujitsu, 2025)
US20250048144A1: Cold-Start Multivariate Time Series Forecasting (Fujitsu, 2025)
(Filed @ Fujitsu Research India, 2022–2025):
Proactive RCA from Log/Text Data with Mamba and LLM
Explainable Timeseries forecasting & Anomaly Detection
Automated Real-Time Storm Chaser System
Causal Graph Linked Prediction for Gene Expression
Enhancing Multivariate Time Series Training Strategies
Anomaly Detection in Timeseries Data
and more (Full list available upon request).
Niraj Kumar, Bhiman Kumar Baghel, “Intent Focused Semantic Parsing and zero-shot Learning for Out-of-Domain detection in Spoken Language Understanding” has been accepted for publication in IEEE Access. [Dec-2021]
Niraj Kumar, Bhiman Kumar Baghel, “Smart Stacking of Deep Learning Models for Granular Joint Intent-Slot Extraction for Multi-Intent SLU” has been accepted for publication in IEEE Access. [June-2021]
Niraj Kumar, Kannan Srinathan and Vasudeva Varma; “Unsupervised Deep Semantic and Logical Analysis for Identification of Solution Posts from Community Answers”; “Int. J. of Information and Decision Sciences”, IJIDS 8(2): 153-178 (2016).
Niraj Kumar, Kannan Srinathan and Vasudeva Varma; “A Graph based Unsupervised N-gram Filtration Technique for Automatic Keyphrase Extraction”; “Int. J. of Data Mining, Modelling and Management”, Vol. 8, No. 2: 124-143, (2016)
Supriya Bajpai, Athira Gopal, Chandrakant, Niraj Kumar; HG-InsightLog: Context Prioritization and Reduction for Question Answering with Non-Natural Language Construct Log Data. Accepted at ACL-2025 Findings.
Supriya Bajpai, Krishna Murthi, Niraj Kumar, AnomGraphAdv: Enhancing Anomaly and Network Intrusion Detection in Wireless Networks Using Adversarial Training and Temporal Graph Networks. Accepted for publication in ACM WiSec 2024.
Nikhil Cherian Kurian, Niraj Kumar; Improved Multi-Step, Multi-Variate, and Spatiotemporal 5G Data Usage Forecasting Without Deploying Data Imputation Techniques; 2023 IEEE Global Communications Conference.
Anant Khandelwal, Niraj Kumar; A Unified System for Aggression Identification in English Code-Mixed and Uni-Lingual Texts. COMAD/CODS 2020: 55-64.
Niraj Kumar; “A Graph Based Automatic Plagiarism Detection Technique to Handle the Artificial Word Reordering and Paraphrasing” CICLing 2014, LNCS 8404, pp. 481–494, 2014. (My work @ TCS Innovation Lab; best paper award, 1st place @ CICLing 2014).
Niraj Kumar and Lipika Dey; “Automatic Quality Assessment of documents with Application to Essay grading”; accepted for publication in MICAI-2013. (My work @ TCS Innovation Lab).
Niraj Kumar, Kannan Srinathan, Vasudeva Varma: A Knowledge Induced Graph-Theoretical Model for Extract and Abstract Single Document Summarization. CICLing (2) 2013: LNCS 7817, pp. 408-423.
Niraj Kumar, Rashmi Gangadharaiah., Kannan Srinathan and Vasudeva Varma; “Exploring the Role of Logically Related Non-Question Phrases for Answering Why-Questions”; Accepted for publication in NLDB-2013.
Niraj Kumar, Kannan Srinathan, and Vasudeva Varma; Using Graph Based Mapping of Co-Occurring Words and Closeness Centrality Score for Summarization Evaluation; A. Gelbukh (Ed.): CICLing 2012, LNCS 7182, pp. 353–365, 2012.
Niraj Kumar, Kannan Srinathan, and Vasudeva Varma; Using Wikipedia Anchor Text and Weighted Clustering Coefficient to Enhance the Traditional Multi-Document Summarization; A. Gelbukh (Ed.): CICLing 2012, LNCS 7182, pp. 390–401, 2012.
Niraj Kumar, Kannan Srinathan, and Vasudeva Varma; Using Unsupervised System with least linguistic features for TAC-AESOP Task; In: Proceedings of Text Analysis Conference (TAC 2011), National Institute of Standards and Technology (NIST), Gaithersburg, Maryland, USA TAC-2011.
Niraj Kumar, Kannan Srinathan, and Vasudeva Varma; An Effective Approach for AESOP and Guided Summarization Task; In: Proceedings of Text Analysis Conference (TAC 2010), National Institute of Standards and Technology (NIST), Gaithersburg, Maryland, USA TAC 2010.
Niraj Kumar, Kannan Srinathan and Vasudeva Varma; Evaluating Information Coverage in Machine Generated Summary and Variable Length Documents; COMAD 2010.
Niraj Kumar, Venkata Vinay Babu Vemula, Kannan Srinathan, Vasudeva Varma: Exploiting N-gram Importance and Wikipedia based Additional Knowledge for Improvements in GAAC based Document Clustering. KDIR 2010: 182-187.
Niraj Kumar, Kannan Srinathan and Vasudeva Varma; Key Fact Extraction from Newswire Articles by Exploiting Local features based weighting and Interaction of sentences,(Published in ICON-2010, length 6-pages)
Niraj Kumar and Kannan Srinathan; A New Approach for Clustering Variable Length Documents,(Published in IEEE IACC-09).
Niraj Kumar, Kannan Srinathan: Automatic keyphrase extraction from scientific documents using N-gram filtration technique. ACM Symposium on Document Engineering 2008: 199-208.